From classrooms to campaigns to code

Looking back, I would have never imagined my current state of career. I graduated from IIT Kanpur with a dual degree in 2015, turned down a campus placement offer from Accenture, and instead left for a funded PhD in Economics at the University of Washington, Seattle. Eight months later I was back in India, and what followed touched six different domains: political campaigns, teaching, data engineering, data science, NLP, generative AI and Legal tech enterpreneurship. All of it happened because I kept asking myself the same question: what do I want to explore next.

Leaving a funded PhD, eight months in

Landing for the first time in a foreign country in first ever flight of life; graduate school started pretty well. I settled into courseworks, interacting with non-Indians for the first time. Too many firsts. I was single Indian in a cohort of 23 people from 15 different countries.

While attending courses, one thing felt very unique, macroeconomics professor taught a very interesting perspective of looking at things: she was teaching everything through their historical evolution, study how a subject actually developed before you study what it looks like today. That’s still how I approach anything unfamiliar.

The trouble started in mid of second quarter while working as a research assistant on Indian government school data. I was working with a chinese professor in a US university on Indian data, and interestingly I was not able to understand the context. Though I lived in India all my life, I felt totally aloof. I could not discriminate - district and village boundaries, how administrative and civil bodies related to each other, how government actually functioned at different levels: none of it was intuitive to me from a classroom in Seattle. I was studying India without having really been inside it, and I could feel the gap.

In parallel I was facing some existential crisis - what actually do I want to achieve in life, what does life mean to me? Would I be able to return India back given limited opportunities in India. In hindsight, I find it very natural, a 23 year old phase of my life. I communicated my thoughts to 2 of my favorite professors there that I would like to return back to India. The head of the university’s statistics department, impressed by a course I had taken with him, offered to bring me into a PhD in statistics instead. I politely turned it down and came home. That decision, and what an unrelated Indian-data research project taught me about context along the way, gets the fuller treatment here: Why I came back from a funded PhD in the US.

IPAC: four campaigns in two years

Back in India, a batchmate mentioned that several IIT Kanpur seniors and juniors were working at IPAC, the Indian Political Action Committee, Prashant Kishor’s political consulting outfit. I applied, at roughly half my PhD stipend, and spent the next two years moving through Punjab, Uttarakhand, Uttar Pradesh, and Andhra Pradesh: reading election data, designing campaign strategy, and then living inside constituencies to run it. That included a stretch in Amritsar during the 2016 surgical strikes, right up against the border.

That period taught me more about how India actually works than anything I had studied: formal power versus informal influence, and how far ground reality can diverge from what the numbers say. The fuller account of those two years is here: What political campaigns taught me about India.

Teaching IIT-JEE, and deciding it wasn’t the path

When I left IPAC, a few of us who had worked closely together, from different IITs and IIMs, started weighing startup ideas. One thread that stuck was education. I was always interested in learning “what is the best way to learn”. Meanwhile a state education minister we had worked with during the Punjab campaign offered support for an ed-tech pilot in government schools. I decided to join a coaching institute in Vadodara and, over about a year and a half, taught IIT-JEE preparatory classes at three national-level institutes to understand the coaching space. Even during my JEE preparation, I was never part of this trend of coachings, and hence there was something very interesting for me to explore.

But by January 2018 I knew ed-tech wasn’t where I wanted to build a career, and I resigned and moved to Pune to figure out what came next.

Into corporate: data engineering, NLP, and a first management role

One mindset I had without any grounding was, ruling out a corporate job as career. Even in introspection I can not find a logic out for that mindset. May be, scarce of inability to secure an internship for second year semester break at IIT Kanpur inflicted a fear of failure. Don’t know the reason.

However, that changed in 2018. I decided to get into corporate job now, interviews were rough at first, since I was involved with statistics or programming in academic setups only for course work and research in IIT Kanpur and UW Seattle. Luckily an IIT Kanpur batchmate vouched for me at his startup and I joined as a data engineer, doing data engineering, data science, and system design works all at once under his guidance.

From there I moved into a Senior Data Scientist role focused on NLP: search built on Elasticsearch and Kibana, spaCy- and NLTK-based pipelines, early embedding models like word2vec and BERT, and the model deployment and dashboarding work that came with all of it. I then joined Bajaj Finserv to optimize search for its mobile app and to lead the analytics team for its Rural Marketplace, my first time managing a team of that size. It was also where I ran into the limits of how much a large organization will invest in deep technical work relative to what the market was starting to reward.

Generative AI, a layoff, and Order.law

As data science field was getting monotonous at some point of time, where people were rarely putting brain and trusting compute too much, I decided to move into a data-engineering-heavy role at a smaller company. And just in time, within three months of joining ChatGPT launched and reshaped the field overnight. I was asked to set up and lead the company’s generative AI effort, but it became clear the initiative was aimed more at investors than at a real product, so I left for a Senior AI Engineer role where they were actually building, fine-tuning, and deploying generative AI models under real data-security and resource constraints, rather than just experimenting to put some slides in investor deck. About a year and a half in, the company hit financial trouble and let go of many of its India engineering team.

Around the same time, a former Bajaj Finserv colleague going through a legal dispute pitched an idea for applying generative AI to legal work. I built a proof of concept in fifteen days, we demoed it to lawyers at the Bombay High Court, and the response was strong enough that we decided to build it properly. That became Order.law, an AI-native legal technology company I co-founded and where I serve as CTO.

Where I am now

I currently lead the AI R&D Lab as Senior AI Architect at Indexnine Technologies, alongside my work at Order.law. Along the way I’ve also worked in NLP, data, and analytics roles at organizations including Simplifai Cognitive Services, Bajaj Finserv, and Avance Consulting.

Looking back

I never had a clear ten-year plan through any of this. I took salary cuts, switched domains, and started over more times than I expected to. But every move left me with context I used to make the next decision: a classroom showed me how people actually learn, an unfinished PhD taught me to distrust data without ground context, four election campaigns showed me how systems really behave versus how they’re supposed to. That instinct is the thread connecting all of it to the AI systems I build today.

I’m still figuring out what comes next. This site is where I’m writing it down as I go.